Image Based Rendering from Perspective and Orthographic Images for Autostereoscopic Multi-View Displays

Abstract

Current autostereoscopic (AS) multi-view displays for video that are targeted at the market allow typically up to 60 frames per second and offer between 20 and 60 different views per pixel. Future full parallax AS displays may well require thousands of views simultaneously. With the large number of different views video displays consume a huge amount of data, either transferred to the display or to be computed on demand from a 3D scene representation. In the following a novel depth-image based rendering interpolation algorithm targeted at multi-view video displays is introduced that combines the results of an interpolation on orthographic and perspective images. The same idea is further utilised to implement an efficient computer graphic rendering algorithm for full parallax AS displays.